Skip to main content

Amazon Bedrock

What is it​

A fully managed service that provides access to high-performance foundation models (FMs) from leading AI companies, as well as Amazon's own models, through a unified API.

What it's for​

Facilitates the building of generative AI applications, allowing developers to experiment, evaluate, and customize FMs with their own data, without the need to manage the underlying infrastructure.

Use cases​

  • Text generation for content creation (articles, emails, blog posts).
  • Summarization of documents and long texts.
  • Code generation for development task automation.
  • Creation of chatbots and virtual assistants with more natural responses.
  • Research and information retrieval from large volumes of data.
  • Image generation from text descriptions.

Key points​

  • FM Access: Provides access to a variety of foundation models, including text, image, and multimodal models.
  • Unified API: Allows interaction with different FMs through a single API.
  • Customization: Enables customization of FMs with your own data using techniques such as fine-tuning and RAG (Retrieval Augmented Generation).
  • Fully managed: AWS manages the underlying infrastructure, allowing developers to focus on building applications.
  • Security and privacy: Your data remains private and is not used to train the foundation models.
  • Model evaluation: Tools to evaluate and compare the performance of different FMs for your use case.

Comparison with developing generative AI models from scratch:​

  • Amazon Bedrock: Simplifies and accelerates the development of generative AI applications by providing access to pre-trained FMs and tools for customization and deployment. Reduces the need for ML expertise and development time.
  • Developing generative AI models from scratch: Requires a team of highly specialized data scientists and ML engineers, large volumes of training data, compute-intensive infrastructure, and significant investment of time and resources. Suitable for very specific use cases that require highly customized and proprietary models.